Sentiment Analysis with Custom Negation Handling

A unique NLP project using logistic regression and TF-IDF

📌 Description

This project performs sentiment analysis on movie reviews using logistic regression. Unlike typical models, it includes custom negation handling to improve accuracy when phrases like “not good” or “didn’t enjoy” are used. This small but powerful change leads to more accurate sentiment predictions.

🚀 Key Features

• Preprocessing (HTML tags, URLs, punctuations removal)
• TF-IDF vectorization with top 5000 features
• Logistic Regression classifier
• Custom handling of negations (e.g., not good → not_good)
• Accuracy comparison: with vs. without negation handling
• Custom test predictions to showcase difference

🧠 Uniqueness

Traditional sentiment analysis often fails to capture the impact of negation. This project introduces a manual negation tagging mechanism, allowing models to better understand phrases like “not happy,” “never liked,” or “don’t recommend.”

📂 Files

• sentiment_with_negation.py: Main project with custom logic
• sentiment_normal.py: Baseline model without negation handling
• README.md: Documentation
• IMDB Dataset.csv: Dataset used for training/testing. Not included due to licensing restrictions. You can download it from Kaggle or other open sources.
• output.txt: Contains comparison of normal vs. custom sentiment outputs, demonstrating negation handling improvements.

License

This project is licensed under the MIT License.
See the LICENSE file for full license details

🔧 How to Run

# Activate virtual environment
.\venv\Scripts\activate

# Install requirements (if you create requirements.txt)
pip install -r requirements.txt

# Run the model
python sentiment_with_negation.py